Stop pasting stack traces into chat

The loop everyone runs today: production breaks, you open the APM, copy a stack trace, paste it into your coding agent, then copy the logs, then the span timings, then the release diff. You are the transfer station between two tools that should be talking to each other.

One install, and the agent rides along

$ npx skills add tracewayapp/traceway

That installs the Traceway skills into Claude Code, Cursor, Codex, or any agent that reads SKILL.md. From there the agent queries Traceway itself:

• pull the exception group, its stack trace, and the 412 events behind it
• read the failing span's logs and the request's full trace
• check the error-rate metric before and after the fix ships

Debugging becomes a conversation

"Fix the checkout 500s" is now a complete instruction. The agent finds TypeError: cart is null, walks the trace to the span that died, reads the session replay context, and proposes the patch — without you ferrying context between tabs.

The skills are in the same MIT repo as everything else. If your agent speaks MCP instead, the MCP server covers the same ground.